摘要
利用FIR滤波器抗噪性能好的特点,设计了一种提取语音特征向量的方法,该方法具有运算量小,运算效率高的特点。运用Matlab作为辅助设计工具,结合VS1003B音频编码芯片与CycloneII系列EP2C8Q208在NIOS II环境下得到了实现。最后通过BP神经网络将得到的特征向量作为样本进行训练,验证通过FIR滤波器提取语音特征向量具备一定的准确性和可识别性。
Using FIR filter with good performance in the aspect of anti-noise, a method of extracting voice feature vectors is designed. The method has light computational load and high efficiency. Using Matlab as auxiliary design tools, combined VS1003B audio codec chip and CyclonelI series EP2C8Q208, the mothod is realized in NIOS II. Through BP neural net- work, vectors are trained as samples to check the method. The way of extracting voice features vectors by FIR filter has accu- racy and identifiability.
出处
《电声技术》
2013年第10期49-53,共5页
Audio Engineering
作者简介
程石磊。硕士研究生，主要研究方向为灾害救助与预警处置系统。
